Intelligent direct-current power supply whole-course fault recording analysis system

Abstract

The invention relates to the technical field of power system monitoring, in particular to an intelligent direct-current power supply whole-course fault recording analysis system which comprises a dynamic threshold self-adaption module, a multi-dimensional data fusion module, an event triggering module, a long-period data accumulation module, a complex boundary condition monitoring module and a fault mode recognition early warning module. Wherein the dynamic threshold adaptive module is used for collecting multi-dimensional data of a direct current power supply system in real time and dynamically adjusting a threshold range of fault detection; the multi-dimensional data fusion module is used for generating a comprehensive system operation state evaluation result; and the event triggering module is used for triggering a fault recording process when an abnormal condition meeting a threshold range condition is detected. According to the invention, through real-time data monitoring, deep learning and fault mode identification and long-term data accumulation, the fault early warning capability and the operation and maintenance efficiency of the DC power supply system are significantly improved, and stable and reliable operation of the system is ensured.

Description

Technical Field

[0001] The invention relates to the technical field of power system monitoring, and in particular to an intelligent direct current power supply full-process fault recording and analysis system. Background Art

[0002] In modern DC power supply systems, especially in important areas such as rail transit, it is crucial to ensure system safety and stability. The common asymptomatic tripping problem in the system, that is, there is no warning when the fault occurs and it is difficult to determine the specific cause afterwards, seriously affects the reliability and efficiency of operations. Current technical solutions mainly rely on simple fault recording equipment and basic fault detection mechanisms. Although these devices can record the system status for a short period of time when a fault occurs, they are limited to instantaneous data before and after the fault occurs. This approach has significant limitations in fault prevention and subsequent analysis, especially when it is difficult to capture key data during the development of the fault, making it difficult to conduct in-depth analysis and timely response to the fault.

[0003] Faced with the limitations of existing technologies, there is an urgent need for a system that can provide comprehensive monitoring and analysis capabilities to enhance the fault diagnosis and prevention capabilities of DC power supply systems in complex environments. The main problems of existing systems include intermittent data recording, limited analysis capabilities, and insufficient response to complex fault modes. These problems pose many challenges to the system operation and maintenance team, such as extended fault handling time, increased operation and maintenance costs, and reduced system reliability.

[0004] Therefore, it is particularly important to develop a system with real-time data monitoring, long-term data accumulation, complex data fusion processing and efficient analysis capabilities. Summary of the invention

[0005] Based on the above purpose, the present invention provides an intelligent DC power supply full-process fault recording and analysis system.

[0006] The intelligent DC power supply full-process fault recording and analysis system includes a dynamic threshold adaptive module, a multi-dimensional data fusion module, an event trigger module, a long-period data accumulation module, a complex boundary condition monitoring module, and a fault mode recognition and early warning module; among which:

[0007] Dynamic threshold adaptive module: used to collect multi-dimensional data of voltage, current, temperature, humidity, and electromagnetic interference of the DC power supply system in real time, and dynamically adjust the threshold range of fault detection based on the preset adaptive threshold adjustment algorithm;

[0008] Multi-dimensional data fusion module: used to receive the threshold range data of fault detection, and conduct comprehensive processing and analysis on various monitoring data of the DC power supply system to generate comprehensive system operation status evaluation results;

[0009] Event trigger module: used to analyze the comprehensive system operation status evaluation results, identify potential abnormal events, and automatically record relevant parameters and trigger the fault recording process when abnormal conditions that meet the threshold range conditions are detected;

[0010] Long-period data accumulation module: used to receive and store relevant parameters and recorded data of the event trigger module, and accumulate historical data of the DC power supply system under normal and abnormal operating conditions, providing data basis for the complex boundary condition monitoring module;

[0011] Complex boundary condition monitoring module: connected to the long-period data accumulation module, using the stored historical data to monitor the state changes of the DC power supply system under critical operating conditions in real time, especially in the case of system load fluctuations and increased external interference, by comparing the current operating data with the historical data, it is determined whether to transmit the boundary state data to the fault mode recognition and early warning module;

[0012] Fault pattern recognition and early warning module: connected to the complex boundary condition monitoring module, used to receive boundary status data, identify potential fault patterns based on deep learning algorithms, and compare the identified patterns with historical data in the long-period data accumulation module. If there are high-risk fault signs, an early warning signal is issued.

[0013] Optionally, the dynamic threshold adaptive module includes a data acquisition unit, a data processing unit and a threshold adjustment unit; wherein:

[0014] Data acquisition unit: used to collect multi-dimensional operating parameters of voltage, current, temperature, humidity and electromagnetic interference in real time through sensors deployed at each predetermined node of the DC power supply system, wherein the sensors include voltage sensors, current sensors, temperature sensors, humidity sensors and electromagnetic interference sensors;

[0015] Data processing unit: used to pre-process the collected multi-dimensional data, including noise removal, data normalization and data formatting;

[0016] Threshold adjustment unit: used to dynamically adjust the threshold range of fault detection based on a preset adaptive threshold adjustment algorithm. The adaptive threshold adjustment algorithm includes a fuzzy logic algorithm and a regression analysis algorithm based on historical data. By analyzing the fault mode in the historical data, the current threshold range is adjusted in real time to adapt to changes in the DC power supply system under different operating conditions.

[0017] Optionally, the threshold adjustment unit includes:

[0018] Historical data extraction: The historical operation data of the DC power supply system is extracted from the long-term data accumulation module, including the parameters of voltage V, current I, temperature T, humidity H and electromagnetic interference E as input variables. The operation data includes the normal operation data of multiple time periods and the data when the fault occurs, which are recorded as V n ,I n ,T n ,H n ,E n and V f ,I f ,T f ,H f ,E f , where n represents the normal state and f represents the fault state;

[0019] Fault mode recognition: Based on the fuzzy logic algorithm, the historical data is processed. First, each input variable V n ,I n ,T n ,H n ,E n and V f ,I f ,T f ,H f ,E f Perform fuzzy processing to convert continuous numerical data into fuzzy sets, which are specifically divided into three fuzzy sets: low voltage, normal voltage, and high voltage;

[0020] Fuzzy rule generation: Based on the historical data after fuzzification, a fuzzy rule base is generated to describe the failure mode of the system under different operating conditions;

[0021] Regression analysis adjustment: Combined with the fuzzy rules output by fuzzy logic, regression analysis is performed based on historical fault data to establish a regression model between input variables V, I, T, H, E and system status Y;

[0022] Real-time threshold adjustment: During the operation of the current system, the regression model obtained in the regression analysis adjustment is used to adjust the real-time collected data V current ,I current ,T current ,H current ,E current Perform analysis and calculate the current system status prediction value like When the threshold value of normal operation is exceeded, the fault detection threshold V is dynamically adjusted. threshold ,V threshold ,T threshold ,H threshold ,E threshold ;

[0023] Threshold range output: The dynamically adjusted fault detection threshold range is passed to the event trigger module as a reference standard for subsequent fault detection.

[0024] Optionally, the multidimensional data fusion module includes a data receiving unit, a data processing unit, a feature extraction unit and a state evaluation unit; wherein:

[0025] Data receiving unit: used to receive the fault detection threshold range data from the dynamic threshold adaptive module and the operation data of the DC power supply system monitored in real time;

[0026] Data processing unit: used to pre-process the received monitoring data, including data cleaning, outlier processing and time synchronization processing;

[0027] Feature extraction unit: used to extract characteristic parameters from the pre-processed monitoring data, including voltage fluctuation rate F V , current fluctuation rate F I , temperature change rate F T , humidity change rate F H And the electromagnetic interference intensity F E ;

[0028] State evaluation unit: used to calculate the comprehensive system operation state evaluation result S based on the extracted characteristic parameters and the threshold range data of fault detection. The formula is: Among them, S is the evaluation result of the system operation status, w V ,w I ,w T ,w H ,w E It is the weight coefficient of each parameter including voltage, current, temperature, humidity and electromagnetic interference.

[0029] Optionally, the event trigger module includes a state analysis unit, an abnormality detection unit and a wave recording control unit; wherein:

[0030] The status analysis unit is used to receive the system operation status evaluation result S from the multi-dimensional data fusion module and perform real-time analysis on the status evaluation result, specifically comparing the evaluation result S with the preset normal operation range [S min , S max ] is compared, if S exceeds the normal operating range, it is determined as a potential abnormal event, and the abnormal event result is transmitted to the abnormality detection unit;

[0031] Anomaly detection unit: used to analyze the system status determined as a potential abnormal event. The anomaly detection unit will decompose the evaluation result S in combination with the current fault detection threshold data to determine whether each monitoring parameter exceeds its corresponding threshold range;

[0032] Recording control unit: used to automatically record relevant system operating parameters and trigger a fault recording process when an abnormal situation that meets the threshold conditions is detected. The fault recording process specifically includes parameter recording, recording startup and data storage.

[0033] Optionally, the long-period data accumulation module includes a data receiving unit, a data storage unit and a data management unit; wherein:

[0034] Data receiving unit: used to receive relevant parameters and recorded data from the event trigger module, including voltage V current 、Current I current , Temperature T current 、Humidity current 、Electromagnetic interference current And the corresponding timestamp information t event ;

[0035] Data storage unit: used to store the received related parameters and recorded data in chronological order; the data storage unit includes a normal operation state data storage subunit and an abnormal operation state data storage subunit;

[0036] Normal operation status data storage subunit: used to store the historical data of the system under normal operation status {V n , I n , T n , H n , E n};

[0037] Abnormal operation state data storage subunit: used to store the recording data of the system under abnormal operation state {V f , I f , T f , H f , E f} and its related timestamp information t f ;

[0038] Data management unit: used to manage and index stored data, including normal operation data and abnormal operation data. The data management unit also includes data retrieval, archiving and backup operations to ensure data integrity and security.

[0039] Optionally, the complex boundary condition monitoring module includes a data comparison unit, a state evaluation unit and a boundary condition judgment unit; wherein:

[0040] Data comparison unit: used to receive the current operation data and the historical data stored in the long-term data accumulation module, and match the current data with the historical data to identify differences or anomalies in operation; the specific comparison methods include:

[0041] Voltage comparison: compare the current voltage V current With the historical voltage V history The voltage change rate difference ΔF is calculated V ;

[0042] Current comparison: compare the current I current With the historical current I history The change trend of the current is calculated, and the current change rate difference ΔF is calculated. I ;

[0043] Temperature comparison: compare the current temperature T current With the historical temperature T history The temperature change rate difference ΔF is calculated T ;

[0044] Humidity comparison: compare the current humidity H current With historical humidity H history The change trend of humidity is calculated, and the humidity change rate difference ΔF is calculated. H ;

[0045] Electromagnetic Interference Comparison: Compare the current electromagnetic interference E current With historical electromagnetic interference E history The changing trend of electromagnetic interference intensity difference ΔF is calculated E ;

[0046] State evaluation unit: connected to the data comparison unit, used to calculate the change rate difference {ΔF V , ΔF I , ΔF T , ΔF H , ΔF E} and the preset critical threshold {V threshold , I threshold , T threshold , H threshold , E threshold} to compare and evaluate whether the current operating state is close to the critical condition. If the difference in the rate of change of one or more parameters exceeds the critical threshold, the system is judged to be in a critical state;

[0047] Boundary condition judgment unit: used to analyze the evaluation results, especially when the system load fluctuates and external interference increases, to determine whether the system needs more in-depth monitoring. Specifically, when the system is in a critical state, the boundary condition judgment unit will determine whether to transmit the boundary state data to the fault mode recognition and warning module based on the comparison results of the current operating data and historical data.

[0048] Optionally, the fault mode recognition and early warning module includes a data receiving unit, a fault mode recognition unit, a historical data comparison unit and an early warning signal generation unit; wherein:

[0049] Data receiving unit: used to receive boundary state data from the complex boundary condition monitoring module, including voltage change rate difference, current change rate difference, temperature change rate difference, humidity change rate difference and electromagnetic interference intensity difference, which are used as input features for fault mode recognition;

[0050] Fault mode identification unit: connected to the data receiving unit, used to analyze the input boundary state data based on the deep learning algorithm, and identify the potential fault modes of the system, including low risk, medium risk and high risk, through the pre-trained neural network model;

[0051] A historical data comparison unit: connected to the fault pattern recognition unit, used to compare the recognized fault pattern with the historical fault data in the long-period data accumulation module, and evaluate the risk level of the current system state by comparing the currently recognized pattern with similar patterns in the historical data;

[0052] Warning signal generation unit: used to generate a warning signal when the historical data comparison unit determines that the system has high-risk failure signs, and send the warning signal to the operation and maintenance personnel interface. The warning signal includes visual and sound alarms.

[0053] Optionally, the fault mode identification unit specifically includes:

[0054] The input layer receives boundary state data: The input layer of the neural network model consists of multiple nodes, each of which corresponds to a characteristic parameter in the received boundary state data. Specifically, the number of input layer nodes is N, and the corresponding received boundary state data is {x1, x1, ..., x N}, where x1 is the voltage change rate difference ΔF V ; x2 is the current change rate difference ΔF I ; x3 is the temperature change rate difference ΔF T ; x4 is the humidity change rate difference ΔF H ; x5 is the electromagnetic interference intensity difference ΔF E ;The input layer is used to normalize all input features and pass them to the hidden layer;

[0055] The hidden layer extracts data features through multi-layer nonlinear mapping: the hidden layer consists of several hidden layers, and the number of nodes in the lth hidden layer is h l , the nodes in the hidden layer process the data in the input layer through a nonlinear activation function to extract the complex patterns and features in the data; the calculation formula of the hidden layer is: in, represents the output of the jth node in the lth layer, represents the output of the i-th node in the previous layer, represents the weight from the i-th node to the j-th node in the l-th layer, is the bias term of the jth node in the lth layer, and f(·) is the activation function;

[0056] The output layer generates the classification results of the fault mode: it receives the features passed by the hidden layer and converts them into specific fault mode classification results. The number of nodes in the output layer corresponds to the number of fault mode classifications of the system. If the number of classifications is set to C, the calculation formula of the output layer is: Among them, y c is the probability value of the failure mode classification result, indicating the probability that the system belongs to the cth failure mode. is the output of the last hidden layer, is the weight from the last hidden layer to the output layer, is the bias term of the output layer, L is the total number of hidden layers, and c represents the classification category; and the output value is converted into a probability value through the softmax function, the formula is: Among them, z i Represents the input of node i in the output layer.

[0057] Optionally, the historical data comparison unit includes:

[0058] Extract historical fault data features: Extract historical fault data related to the currently identified fault mode from the long-term data accumulation module. The data includes the historical fault mode feature vector H k ={h k1 ,h k2 , ..., h kn}, where k represents the number of the historical fault event, h ki represents the i-th feature associated with the fault event;

[0059] Calculate the feature vector of the current fault mode: Based on the fault mode identified by the current system, calculate its corresponding feature vector F = {f1, f2, ..., f n}, where f iRepresents the value of each characteristic parameter in the current fault mode;

[0060] Similarity calculation: Use the cosine similarity algorithm to calculate the current fault mode feature vector F and each historical fault mode feature vector H k The similarity between k The calculation formula is:

[0061] Among them, F.H. k represents the vector inner product, ||F|| and ||H k || respectively vectors F and H k The modulus length; similarity S k The value range of is between [-1, 1]. The closer the value is to 1, the more similar the current fault mode is to the historical mode.

[0062] Risk level assessment: Based on the calculated similarity S k , evaluate the risk level of the current system state, set the similarity threshold θ, if max(S k ) exceeds the threshold θ, the current fault mode is judged to be similar to the high-risk historical mode, and the system is judged to have a high risk; if max(S k ) is close to 1 and exceeds the threshold θ, the risk level is high; if max(S k ) fluctuates around θ, the risk level is medium; if max(S k ) is less than θ, the risk level is low;

[0063] Output risk assessment results: The assessed risk level results are transmitted to the early warning signal generation unit. If the assessment result is high or medium, a corresponding early warning signal is generated to remind the system operator to take corresponding measures; if the risk level is low, the result is recorded but no early warning signal is triggered.

[0064] Beneficial effects of the present invention:

[0065] The present invention significantly improves the monitoring and fault handling capabilities of the DC power supply system by integrating functions such as multi-dimensional data fusion, long-term data accumulation, real-time monitoring and fault warning. First, the system uses a dynamic threshold adaptive module and a multi-dimensional data fusion module to collect and analyze various operating data of the DC power supply system in real time, such as voltage, current, temperature, etc. This not only enhances the early warning capability of impending faults, but also enables operation and maintenance personnel to take measures before the fault occurs, thereby reducing system downtime and maintenance costs.

[0066] The present invention, through the design of a long-period data accumulation module and a complex boundary condition monitoring module, can store a large amount of historical data, and on this basis, conduct an in-depth analysis of the long-term operating status of the system, which enables the system to not only handle immediate abnormal situations, but also learn and predict possible problems in the future. Combined with the deep learning algorithm used in the fault pattern recognition and early warning module, it can identify complex fault patterns from historical data, issue early warnings for potential high-risk events, and greatly improve the reliability and safety of the entire power supply system. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0068] Figure 1 Schematic diagram of a DC power supply full-process fault recording and analysis system according to an embodiment of the present invention;

[0069] Figure 2 Schematic diagram of a fault mode recognition and early warning module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0071] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0072] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0073] like Figure 1-Figure 2 As shown, the intelligent DC power supply full-process fault recording and analysis system includes a dynamic threshold adaptive module, a multi-dimensional data fusion module, an event trigger module, a long-period data accumulation module, a complex boundary condition monitoring module, and a fault mode recognition and early warning module; wherein:

[0074] Dynamic threshold adaptive module: used to collect multi-dimensional data of voltage, current, temperature, humidity, and electromagnetic interference of the DC power supply system in real time, and dynamically adjust the threshold range of fault detection based on the preset adaptive threshold adjustment algorithm, and send the adjusted threshold data to the multi-dimensional data fusion module;

[0075] Multi-dimensional data fusion module: used to receive the threshold range data of fault detection, and conduct comprehensive processing and analysis on various monitoring data of the DC power supply system, generate a comprehensive system operation status evaluation result, and send the evaluation result to the event trigger module;

[0076] Event trigger module: used to analyze the comprehensive system operation status evaluation results, identify potential abnormal events, and automatically record relevant parameters and trigger the fault recording process when an abnormal situation that meets the threshold range conditions is detected, and transmit the relevant parameters and recording data to the long-term data accumulation module;

[0077] Long-period data accumulation module: used to receive and store relevant parameters and recorded data of the event trigger module, and accumulate historical data of the DC power supply system under normal and abnormal operating conditions, providing data basis for the complex boundary condition monitoring module;

[0078] Complex boundary condition monitoring module: connected with the long-period data accumulation module, it uses the stored historical data to monitor the state changes of the DC power supply system under critical operating conditions in real time, especially when the system load fluctuates and external interference increases. By comparing the current operating data with the historical data, it determines whether to transmit the boundary state data to the fault mode recognition and early warning module;

[0079] Fault pattern recognition and early warning module: connected to the complex boundary condition monitoring module, used to receive boundary status data, identify potential fault patterns based on deep learning algorithms, and compare the identified patterns with historical data in the long-term data accumulation module. If there are high-risk fault signs, a warning signal will be issued.

[0080] The dynamic threshold adaptive module includes a data acquisition unit, a data processing unit and a threshold adjustment unit; wherein:

[0081] Data acquisition unit: used to collect multi-dimensional operating parameters such as voltage, current, temperature, humidity and electromagnetic interference in real time through sensors deployed at each predetermined node of the DC power supply system. The sensors include voltage sensors, current sensors, temperature sensors, humidity sensors and electromagnetic interference sensors. Each sensor is connected to the data acquisition unit to ensure that the collected data is complete and can be updated in real time;

[0082] Data processing unit: used to pre-process the collected multi-dimensional data, including noise removal, data normalization and data formatting. Noise removal is used to eliminate high-frequency noise introduced by sensor errors or external interference in the collected data, thereby improving the accuracy of the data; data normalization is used to convert the collected data to a unified scale range to ensure that data of different dimensions are comparable in subsequent analysis and avoid calculation errors caused by dimensional differences; data formatting is used to convert the pre-processed data into a standardized format that the system can recognize and process, ensuring consistency and integrity of the data during transmission and use between modules;

[0083] Threshold adjustment unit: used to dynamically adjust the threshold range of fault detection based on a preset adaptive threshold adjustment algorithm. The adaptive threshold adjustment algorithm includes a fuzzy logic algorithm and a regression analysis algorithm based on historical data. By analyzing the fault modes in the historical data, the current threshold range is adjusted in real time to adapt to changes in the DC power supply system under different operating conditions, thereby ensuring the accuracy of fault detection.

[0084] The threshold adjustment unit comprises:

[0085] Historical data extraction: The historical operation data of the DC power supply system is extracted from the long-term data accumulation module, including the parameters of voltage V, current I, temperature T, humidity H and electromagnetic interference E as input variables. The operation data includes the normal operation data of multiple time periods and the data when the fault occurs, which are recorded as V n ,I n ,T n ,H n ,E n and V f ,I f ,T f,H f ,E f , where n represents the normal state and f represents the fault state;

[0086] Fault mode recognition: Based on the fuzzy logic algorithm, the historical data is processed. First, each input variable V n ,I n ,T n ,H n ,E n and V f ,I f ,T f ,H f ,E f Fuzzy processing is performed to convert continuous numerical data into fuzzy sets, which are specifically divided into three fuzzy sets: low voltage, normal voltage, and high voltage, expressed as:

[0087]

[0088] Among them, μ L (V) is the membership function of voltage V belonging to the low voltage set, V L1 and V L2 The threshold for voltage division;

[0089] Fuzzy rule generation: Based on the fuzzy processed historical data, a fuzzy rule base is generated to describe the failure mode of the system under different operating conditions; for example: Rule 1: If the voltage is low voltage and the current is high current, then the system state is failure. The weights of various rules are determined by the occurrence frequency of different failure modes in the historical data;

[0090] Regression analysis adjustment: Combined with the fuzzy rules output by fuzzy logic, regression analysis is performed based on historical fault data to establish a regression model between input variables V, I, T, H, E and system state Y. The expression is: Y = β0 + β V V+β I I+β T T+β H H+β E E+∈, where Y represents the system state, β0 is a constant term, and β V ,β I ,β T ,β H ,β E are the regression coefficients of voltage, current, temperature, humidity and electromagnetic interference respectively, ∈ is the error term, and the regression coefficient β is solved by the least squares method V ,β I ,β T ,β H ,β E , and obtain the optimal regression model;

[0091] Real-time threshold adjustment: During the operation of the current system, the regression model obtained in the regression analysis adjustment is used to adjust the real-time collected data V current ,I current ,T current ,H current ,E current Perform analysis and calculate the current system status prediction value The formula is: like When the threshold value of normal operation is exceeded, the fault detection threshold V is dynamically adjusted. threshold ,I threshold ,T threshold ,H threshold ,E threshold , ensuring that the system has sufficient sensitivity for fault detection under different operating conditions;

[0092] Threshold range output: The dynamically adjusted fault detection threshold range is passed to the event trigger module as a reference standard for subsequent fault detection, ensuring that the system can respond quickly according to real-time operating conditions; through the above steps, the threshold adjustment unit can accurately adjust the fault detection threshold of the DC power supply system based on historical fault modes and changes in real-time data, ensuring that the system can respond to potential fault risks in a timely manner under different operating environments.

[0093] The multi-dimensional data fusion module includes a data receiving unit, a data processing unit, a feature extraction unit and a state evaluation unit; wherein:

[0094] Data receiving unit: used to receive the fault detection threshold range data from the dynamic threshold adaptive module and the operation data of the DC power supply system monitored in real time;

[0095] Data processing unit: used to pre-process the received monitoring data, including data cleaning, outlier processing and time synchronization processing; data cleaning is used to remove noise and interference signals in the monitoring data to ensure the accuracy of the data; outlier processing is used to identify and correct outliers in the monitoring data to prevent the outliers from affecting the accuracy of the system evaluation; time synchronization processing is used to synchronize various types of monitoring data according to a unified timestamp to ensure the time consistency between different data sources;

[0096] Feature extraction unit: used to extract characteristic parameters from the pre-processed monitoring data, including voltage fluctuation rate F V , current fluctuation rate F I , temperature change rate F T , humidity change rate F H And the electromagnetic interference intensity FE , the specific formula is as follows: Among them, ΔV, ΔI, ΔT, ΔH, and ΔE represent the changes of voltage, current, temperature, humidity, and electromagnetic interference within the time interval Δt, respectively;

[0097] State evaluation unit: used to calculate the comprehensive system operation state evaluation result S based on the extracted characteristic parameters and the threshold range data of fault detection. The formula is: Among them, S is the evaluation result of the system operation status, w V ,w I ,w T ,w H ,w E are the weight coefficients of the voltage, current, temperature, humidity, and electromagnetic interference parameters, which are determined by analyzing historical data; the comprehensive status result S is transmitted to the event trigger module to determine whether fault processing or further recording analysis is required; through the above steps, the multi-dimensional data fusion module can effectively integrate and analyze the multi-dimensional monitoring data of the DC power supply system, and generate accurate system status evaluation results according to the current operating conditions, providing a reliable basis for subsequent fault detection and processing.

[0098] The event trigger module includes a state analysis unit, an abnormality detection unit and a recording control unit; wherein:

[0099] The status analysis unit is used to receive the system operation status evaluation result S from the multi-dimensional data fusion module and perform real-time analysis on the status evaluation result, specifically comparing the evaluation result S with the preset normal operation range [S min , S max ] is compared, if S exceeds the normal operating range, it is determined as a potential abnormal event, and the abnormal event result is transmitted to the abnormality detection unit;

[0100] Abnormal detection unit: used to analyze the system status determined as a potential abnormal event. The abnormal detection unit will decompose the evaluation result S in combination with the current fault detection threshold data to determine whether each monitoring parameter exceeds its corresponding threshold range; the specific analysis includes: voltage abnormality detection, determining the voltage characteristic parameter F in S V Does it satisfy F? V >V threshold ; Current abnormality detection, judging current characteristic parameter F I Does it satisfy F? I >I threshold ; Temperature anomaly detection, determine the temperature characteristic parameter F T Does it satisfy F? T >T threshold ; Humidity anomaly detection, determine humidity characteristic parameter F H Does it satisfy F?H >H threshold ; Detect electromagnetic interference anomalies and determine the electromagnetic interference characteristic parameters F E Does it satisfy F? E >E threshold ; If the test result shows that one or more parameters exceed the threshold range, the abnormality detection unit determines that an abnormal event exists and transmits the relevant data to the recording control unit;

[0101] Recording control unit: used to automatically record relevant system operating parameters and trigger the fault recording process when an abnormal situation that meets the threshold conditions is detected. The fault recording process specifically includes parameter recording, recording start and data storage;

[0102] Parameter recording is used to record all parameter data related to the abnormal event immediately after the abnormal event is confirmed, including voltage V current 、Current I current , Temperature T current 、Humidity current and electromagnetic interference E current ;

[0103] Recording start is used to start the fault recording device, continuously recording all system data before and after the abnormal event occurs, ensuring a complete record of the fault process;

[0104] Data storage is used to store the recorded data together with the timestamp of the abnormal event, the detection results and other information in the long-period data accumulation module for subsequent analysis; through the above description, the event trigger module can effectively analyze the comprehensive system operation status evaluation result S, and when an abnormal situation that meets the threshold conditions is detected, it automatically records the relevant parameters and triggers the fault recording process, thereby providing detailed data support for subsequent fault analysis.

[0105] The long-period data accumulation module includes a data receiving unit, a data storage unit and a data management unit; wherein:

[0106] Data receiving unit: used to receive relevant parameters and recorded data from the event trigger module, including voltage V current 、Current I current , Temperature T current 、Humidity current 、Electromagnetic interference current And the corresponding timestamp information t event ,This unit ensures that all received data are associated with the abnormal event detection results in the event trigger module, facilitating subsequent data processing and storage;

[0107] Data storage unit: used to store the received related parameters and recorded data in chronological order; the data storage unit includes a normal operation state data storage subunit and an abnormal operation state data storage subunit;

[0108] Normal operation status data storage subunit: used to store the historical data of the system under normal operation status {V n , I n , T n , H n , E n}, these data are mainly used to establish the baseline of system operation and provide comparative analysis for the complex boundary condition monitoring module;

[0109] Abnormal operation state data storage subunit: used to store the recording data of the system under abnormal operation state {V f , I f , T f , H f , E f} and its related timestamp information t f ,These data are important basis for subsequent fault mode identification;

[0110] Data management unit: used to manage and index stored data, including normal operation data and abnormal operation data. The data management unit also includes data retrieval, archiving and backup operations to ensure data integrity and security;

[0111] The specific operations are as follows:

[0112] Data classification management: Data is classified and stored according to the operating status (normal or abnormal), and each data set is assigned a unique identifier for quick retrieval;

[0113] Data update and archiving: Update the stored data regularly and archive the historical data that exceeds a certain period of time to ensure the effective use of storage space;

[0114] Data backup: Back up key data regularly to prevent data loss or damage and ensure rapid recovery in the event of data failure. Through the above units, the long-period data accumulation module can effectively receive and store the relevant parameters and recorded data of the event trigger module, and accumulate historical data of the DC power supply system under normal and abnormal operating conditions through systematic data management. These data provide a solid data foundation for the complex boundary condition monitoring module and support the continuous optimization of the system and fault warning analysis.

[0115] The complex boundary condition monitoring module includes a data comparison unit, a state evaluation unit and a boundary condition judgment unit; wherein:

[0116] Data comparison unit: used to receive the current operation data and the historical data stored in the long-term data accumulation module, and match the current data with the historical data to identify differences or anomalies in operation; the specific comparison methods include:

[0117] Voltage comparison: compare the current voltage V current With the historical voltage V history The voltage change rate difference ΔF is calculated V , the formula is: Among them, F V Indicates the rate of voltage change;

[0118] Current comparison: compare the current I current With the historical current I history The change trend of the current is calculated, and the current change rate difference ΔF is calculated. I , the formula is: Among them, F I Indicates the rate of change of current;

[0119] Temperature comparison: compare the current temperature T current With the historical temperature T history The temperature change rate difference ΔF is calculated T , the formula is: Among them, F T Indicates the rate of temperature change;

[0120] Humidity comparison: compare the current humidity H current With historical humidity H history The change trend of humidity is calculated, and the humidity change rate difference ΔF is calculated. H , the formula is: Among them, F H Indicates the rate of change of humidity;

[0121] Electromagnetic interference comparison: compare the current electromagnetic interference H current With historical electromagnetic interference E history The changing trend of electromagnetic interference intensity difference ΔF is calculated E , the formula is: Among them, F E Indicates the intensity of electromagnetic interference;

[0122] State evaluation unit: connected to the data comparison unit, used to calculate the change rate difference {ΔF V , ΔF I , ΔF T , ΔF H , ΔF E} and the preset critical threshold {V threshold , I threshold , T threshold , Hthreshold , E threshold} to compare and evaluate whether the current operating state is close to the critical condition. If the difference in the rate of change of one or more parameters exceeds the critical threshold, the system is determined to be in a critical state and the evaluation result is transmitted to the boundary condition judgment unit;

[0123] Boundary condition judgment unit: used to analyze the evaluation results, especially to judge whether the system needs more in-depth monitoring when the system load fluctuates and external interference increases. Specifically, when the system is in a critical state, the boundary condition judgment unit will judge whether to transmit the boundary state data to the fault mode recognition and warning module based on the comparison results of the current operating data and historical data. If the boundary condition judgment unit confirms that the system has a high risk of failure, the data will be immediately transmitted to the fault mode recognition and warning module to trigger further fault identification and warning measures. Through the above description, the complex boundary condition monitoring module can use the stored historical data to monitor the state changes of the DC power supply system under critical operating conditions in real time. Especially when the system load fluctuates and external interference increases, the module can judge whether it is necessary to transmit the boundary state data to the fault mode recognition and warning module in a timely manner through precise comparison and evaluation, thereby providing reliable support for the system's fault warning and safe operation.

[0124] The fault mode recognition and early warning module includes a data receiving unit, a fault mode recognition unit, a historical data comparison unit and an early warning signal generation unit; wherein:

[0125] Data receiving unit: used to receive boundary state data from the complex boundary condition monitoring module, including voltage change rate difference, current change rate difference, temperature change rate difference, humidity change rate difference and electromagnetic interference intensity difference, which are used as input features for fault mode identification and further analysis;

[0126] Fault mode recognition unit: connected to the data receiving unit, used to analyze the input boundary state data based on the deep learning algorithm, and identify the potential fault mode of the system through the pre-trained neural network model. The input layer of the neural network model receives the boundary state data, and the hidden layer extracts the complex patterns and features in the data through multi-layer nonlinear mapping, and finally generates the classification results of the fault mode in the output layer. This classification result includes low risk, medium risk and high risk;

[0127] Historical data comparison unit: connected to the fault pattern recognition unit, used to compare the identified fault pattern with the historical fault data in the long-period data accumulation module, and evaluate the risk level of the current system state by comparing the currently identified pattern with similar patterns in the historical data. If the currently identified pattern is highly similar to the high-risk fault pattern in the historical data, it is determined that the system has a high fault risk;

[0128] Warning signal generation unit: used to generate a warning signal when the historical data comparison unit determines that the system has high-risk fault signs, and send the warning signal to the operation and maintenance personnel interface so that preventive or corrective measures can be taken in time. The warning signal includes visual and sound alarms; through the above unit, the fault mode recognition and warning module can efficiently identify potential fault modes based on deep learning algorithms after receiving boundary status data, and compare the identified modes with historical data. When there are high-risk fault signs, the system will automatically generate a warning signal to provide timely and effective protection measures for the safe operation of the DC power supply system.

[0129] The fault mode identification unit specifically includes:

[0130] The input layer receives boundary state data: The input layer of the neural network model consists of multiple nodes, each of which corresponds to a characteristic parameter in the received boundary state data. Specifically, the number of input layer nodes is N, and the corresponding received boundary state data is {x1, x1, ..., x N}, where x1 is the voltage change rate difference ΔF V ; x2 is the current change rate difference ΔF I ; x3 is the temperature change rate difference ΔF T ; x4 is the humidity change rate difference ΔF H ; x5 is the electromagnetic interference intensity difference ΔF E ; The input layer is used to standardize all input features and pass them to the hidden layer. The formula for standardization is: Among them, x i ′ is the standardized input feature, and They are the features x i The mean and standard deviation of

[0131] The hidden layer extracts data features through multi-layer nonlinear mapping: the hidden layer consists of several hidden layers, the number of nodes in each hidden layer can be different, and the number of nodes in the lth hidden layer is h l , the nodes in the hidden layer process the data in the input layer through a nonlinear activation function to extract the complex patterns and features in the data; the calculation formula of the hidden layer is: in, represents the output of the jth node in the lth layer, represents the output of the i-th node in the previous layer, represents the weight from the i-th node to the j-th node in the l-th layer, is the bias term of the jth node in the lth layer, f(·) is the activation function (such as ReLU function or Sigmoid function), which is usually defined as: f(x) = max(0, x), (ReLU function);

[0132] The output layer generates the classification results of the fault mode: it receives the features passed by the hidden layer and converts them into specific fault mode classification results. The number of nodes in the output layer corresponds to the number of fault mode classifications of the system. If the number of classifications is set to C, the calculation formula of the output layer is: Among them, y c is the probability value of the failure mode classification result, indicating the probability that the system belongs to the cth failure mode. is the output of the last hidden layer, is the weight from the last hidden layer to the output layer, is the bias term of the output layer, L is the total number of hidden layers, and c represents the classification category; and the output value is converted into a probability value through the softmax function, the formula is: Among them, z i represents the input of node i in the output layer; the classification result of the final output layer divides the fault mode into low risk, medium risk or high risk categories, and determines the possible current fault mode of the system based on these probability values, and passes the result to the historical data comparison unit for further processing; through the above steps, the fault mode recognition unit can effectively use the neural network model to extract complex features in the input data, and generate accurate fault mode classification results through multi-layer nonlinear mapping, providing strong support for the system's fault warning.

[0133] The historical data comparison unit includes:

[0134] Extract historical fault data features: Extract historical fault data related to the currently identified fault mode from the long-term data accumulation module. The data includes the historical fault mode feature vector H k ={h k1 ,h k2 , ..., h kn}, where k represents the number of the historical fault event, h ki Represents the i-th feature related to the fault event (such as the rate of change of parameters such as voltage and current);

[0135] Calculate the feature vector of the current fault mode: Based on the fault mode identified by the current system, calculate its corresponding feature vector F = {f1, f2, ..., f n}, where f i Represents the value of each characteristic parameter (such as voltage change rate, current change rate, etc.) in the current fault mode;

[0136] Similarity calculation: Use the cosine similarity algorithm to calculate the current fault mode feature vector F and each historical fault mode feature vector H k The similarity betweenk The calculation formula is:

[0137] Among them, F.H. k represents the vector inner product, ||F|| and ||H k || respectively vectors F and H k The modulus length; similarity S k The value range of is between [-1, 1]. The closer the value is to 1, the more similar the current fault mode is to the historical mode.

[0138] Risk level assessment: Based on the calculated similarity S k , evaluate the risk level of the current system state, set the similarity threshold θ, if max(S k ) exceeds the threshold θ, the current fault mode is judged to be similar to the high-risk historical mode, and the system is judged to have a high risk; if max(S k ) is close to 1 and exceeds the threshold θ, the risk level is high; if max(S k ) fluctuates around θ, the risk level is medium; if max(S k ) is less than θ, the risk level is low;

[0139] Output risk assessment results: The assessed risk level results are passed to the early warning signal generation unit. If the assessment result is high or medium, a corresponding early warning signal is generated to remind the system operator to take corresponding measures; if the risk level is low, the result is recorded but the early warning signal is not triggered; through the above steps, the fault pattern recognition unit can accurately compare the currently identified fault mode with the historical data, and determine the risk of the current system state by calculating the similarity and assessing the risk level, ensuring that an effective early warning is issued before a potential fault occurs.

[0140] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0141] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. Intelligent DC power supply full-process fault recording and analysis system, It is characterized in that It includes dynamic threshold adaptive module, multi-dimensional data fusion module, event trigger module, long-period data accumulation module, complex boundary condition monitoring module and fault mode recognition and early warning module; among which: Dynamic threshold adaptive module: used to collect multi-dimensional data of voltage, current, temperature, humidity, and electromagnetic interference of the DC power supply system in real time, and dynamically adjust the threshold range of fault detection based on the preset adaptive threshold adjustment algorithm; Multi-dimensional data fusion module: used to receive the threshold range data of fault detection, and conduct comprehensive processing and analysis on various monitoring data of the DC power supply system to generate comprehensive system operation status evaluation results; Event trigger module: used to analyze the comprehensive system operation status evaluation results, identify potential abnormal events, and automatically record relevant parameters and trigger the fault recording process when abnormal conditions that meet the threshold range conditions are detected; Long-period data accumulation module: used to receive and store relevant parameters and recorded data of the event trigger module, and accumulate historical data of the DC power supply system under normal and abnormal operating conditions, providing data basis for the complex boundary condition monitoring module; Complex boundary condition monitoring module: connected to the long-period data accumulation module, using the stored historical data to monitor the state changes of the DC power supply system under critical operating conditions in real time, especially in the case of system load fluctuations and increased external interference, by comparing the current operating data with the historical data, it is determined whether to transmit the boundary state data to the fault mode recognition and early warning module; Fault pattern recognition and early warning module: connected to the complex boundary condition monitoring module, used to receive boundary status data, identify potential fault patterns based on deep learning algorithms, and compare the identified patterns with historical data in the long-period data accumulation module. If there are high-risk fault signs, an early warning signal is issued.

2. According to claim 1, the intelligent DC power supply full-process fault recording and analysis system, It is characterized in that The dynamic threshold adaptive module includes a data acquisition unit, a data processing unit and a threshold adjustment unit; wherein: Data acquisition unit: used to collect multi-dimensional operating parameters of voltage, current, temperature, humidity and electromagnetic interference in real time through sensors deployed at each predetermined node of the DC power supply system, wherein the sensors include voltage sensors, current sensors, temperature sensors, humidity sensors and electromagnetic interference sensors; Data processing unit: used to pre-process the collected multi-dimensional data, including noise removal, data normalization and data formatting; Threshold adjustment unit: used to dynamically adjust the threshold range of fault detection based on a preset adaptive threshold adjustment algorithm. The adaptive threshold adjustment algorithm includes a fuzzy logic algorithm and a regression analysis algorithm based on historical data. By analyzing the fault mode in the historical data, the current threshold range is adjusted in real time to adapt to changes in the DC power supply system under different operating conditions.

3. According to claim 2, the intelligent DC power supply full-process fault recording and analysis system, It is characterized in that The threshold adjustment unit comprises: Historical data extraction: The historical operation data of the DC power supply system is extracted from the long-term data accumulation module, including the parameters of voltage V, current I, temperature T, humidity H and electromagnetic interference E as input variables. The operation data includes the normal operation data of multiple time periods and the data when the fault occurs, which are recorded as V n ,I n ,T n ,H n ,E n and V f ,I f ,T f ,H f ,E f , where n represents the normal state and f represents the fault state; Fault mode recognition: Based on the fuzzy logic algorithm, the historical data is processed. First, each input variable V n ,I n ,T n ,H n ,E n and V f ,I f ,T f ,H f ,E f Perform fuzzy processing to convert continuous numerical data into fuzzy sets, which are specifically divided into three fuzzy sets: low voltage, normal voltage, and high voltage; Fuzzy rule generation: Based on the historical data after fuzzification, a fuzzy rule base is generated to describe the failure mode of the system under different operating conditions; Regression analysis adjustment: Combined with the fuzzy rules output by fuzzy logic, regression analysis is performed based on historical fault data to establish a regression model between input variables V, I, T, H, E and system status Y; Real-time threshold adjustment: During the operation of the current system, the regression model obtained in the regression analysis adjustment is used to adjust the real-time collected data V current ,I current ,T current ,H current ,E current Perform analysis and calculate the current system status prediction value like When the threshold value of normal operation is exceeded, the fault detection threshold V is dynamically adjusted. threshold ,I threshold ,T threshold ,H threshold ,E threshold ; Threshold range output: The dynamically adjusted fault detection threshold range is passed to the event trigger module as a reference standard for subsequent fault detection.

4. According to claim 1, the intelligent DC power supply full-process fault recording and analysis system, It is characterized in that The multi-dimensional data fusion module includes a data receiving unit, a data processing unit, a feature extraction unit and a state evaluation unit; wherein: Data receiving unit: used to receive the fault detection threshold range data from the dynamic threshold adaptive module and the operation data of the DC power supply system monitored in real time; Data processing unit: used to pre-process the received monitoring data, including data cleaning, outlier processing and time synchronization processing; Feature extraction unit: used to extract characteristic parameters from the pre-processed monitoring data, including voltage fluctuation rate F V , current fluctuation rate F I , temperature change rate F T , humidity change rate F H And the electromagnetic interference intensity F E ; State evaluation unit: used to calculate the comprehensive system operation state evaluation result S based on the extracted characteristic parameters and the threshold range data of fault detection. The formula is: Among them, S is the evaluation result of the system operation status, w V ,w I ,w T ,w H ,w E It is the weight coefficient of each parameter including voltage, current, temperature, humidity and electromagnetic interference.

5. According to claim 4, the intelligent DC power supply full-process fault recording and analysis system, It is characterized in that The event trigger module includes a state analysis unit, an abnormality detection unit and a wave recording control unit; wherein: The status analysis unit is used to receive the system operation status evaluation result S from the multi-dimensional data fusion module and perform real-time analysis on the status evaluation result, specifically comparing the evaluation result S with the preset normal operation range [S min , S max ] is compared, if S exceeds the normal operating range, it is determined as a potential abnormal event, and the abnormal event result is transmitted to the abnormality detection unit; Anomaly detection unit: used to analyze the system status determined as a potential abnormal event. The anomaly detection unit will decompose the evaluation result S in combination with the current fault detection threshold data to determine whether each monitoring parameter exceeds its corresponding threshold range; Recording control unit: used to automatically record relevant system operating parameters and trigger a fault recording process when an abnormal situation that meets the threshold conditions is detected. The fault recording process specifically includes parameter recording, recording startup and data storage.

6. The intelligent DC power supply full-process fault recording and analysis system according to claim 1, It is characterized in that The long-period data accumulation module includes a data receiving unit, a data storage unit and a data management unit; wherein: Data receiving unit: used to receive relevant parameters and recorded data from the event trigger module, including voltage V current 、Current I current , Temperature T current 、Humidity current 、Electromagnetic interference E current And the corresponding timestamp information t event ; Data storage unit: used to store the received related parameters and recorded data in chronological order; the data storage unit includes a normal operation state data storage subunit and an abnormal operation state data storage subunit; Normal operation status data storage subunit: used to store the historical data of the system under normal operation status {V n , I n , T n , H n , E n }; Abnormal operation state data storage subunit: used to store the recording data of the system under abnormal operation state {V f , I f , T f , H f , E f } and its related timestamp information t f ; Data management unit: used to manage and index stored data, including normal operation data and abnormal operation data. The data management unit also includes data retrieval, archiving and backup operations to ensure data integrity and security.

7. The intelligent DC power supply full-process fault recording and analysis system according to claim 1, It is characterized in that The complex boundary condition monitoring module includes a data comparison unit, a state evaluation unit and a boundary condition judgment unit; wherein: Data comparison unit: used to receive the current operation data and the historical data stored in the long-term data accumulation module, and match the current data with the historical data to identify differences or anomalies in operation; the specific comparison methods include: Voltage comparison: compare the current voltage V current With the historical voltage V history The voltage change rate difference ΔF is calculated V ; Current comparison: compare the current I current With the historical current I history The change trend of the current is calculated, and the current change rate difference ΔF is calculated. I ; Temperature comparison: compare the current temperature T current With the historical temperature T history The temperature change rate difference ΔF is calculated T ; Humidity comparison: compare the current humidity H current With historical humidity H history The change trend of humidity is calculated, and the humidity change rate difference ΔF is calculated. H ; Electromagnetic Interference Comparison: Compare the current electromagnetic interference E current With historical electromagnetic interference E history The changing trend of electromagnetic interference intensity difference ΔF is calculated E ; State evaluation unit: connected to the data comparison unit, used to calculate the change rate difference {ΔF V , ΔF I , ΔF T , ΔF H , ΔF E } and the preset critical threshold {V threshold , I threshold , T threshold , H threshold , E threshold ] to compare and evaluate whether the current operating state is close to the critical condition. If the difference in the rate of change of one or more parameters exceeds the critical threshold, the system is judged to be in a critical state. Boundary condition judgment unit: used to analyze the evaluation results, especially when the system load fluctuates and external interference increases, to determine whether the system needs more in-depth monitoring. Specifically, when the system is in a critical state, the boundary condition judgment unit will determine whether to transmit the boundary state data to the fault mode recognition and warning module based on the comparison results of the current operating data and historical data.

8. The intelligent DC power supply full-process fault recording and analysis system according to claim 1, It is characterized in that The fault mode recognition and early warning module includes a data receiving unit, a fault mode recognition unit, a historical data comparison unit and an early warning signal generation unit; wherein: Data receiving unit: used to receive boundary state data from the complex boundary condition monitoring module, including voltage change rate difference, current change rate difference, temperature change rate difference, humidity change rate difference and electromagnetic interference intensity difference, which are used as input features for fault mode recognition; Fault mode identification unit: connected to the data receiving unit, used to analyze the input boundary state data based on the deep learning algorithm, and identify the potential fault modes of the system, including low risk, medium risk and high risk, through the pre-trained neural network model; A historical data comparison unit: connected to the fault pattern recognition unit, used to compare the identified fault pattern with the historical fault data in the long-period data accumulation module, and evaluate the risk level of the current system state by comparing the currently identified pattern with similar patterns in the historical data; Warning signal generation unit: used to generate a warning signal when the historical data comparison unit determines that the system has high-risk failure signs, and send the warning signal to the operation and maintenance personnel interface. The warning signal includes visual and sound alarms.

9. The intelligent DC power supply full-process fault recording and analysis system according to claim 7, It is characterized in that The fault mode identification unit specifically includes: The input layer receives boundary state data: The input layer of the neural network model consists of multiple nodes, each of which corresponds to a characteristic parameter in the received boundary state data. Specifically, the number of input layer nodes is N, and the corresponding received boundary state data is {x 1 , x 1 , ..., x N }, where x 1 is the voltage change rate difference ΔF V ;x 2 is the current change rate difference ΔF I ;x 3 is the temperature change rate difference ΔF T ;x 4 is the humidity change rate difference ΔF H ;x 5 is the electromagnetic interference intensity difference ΔF E ;The input layer is used to normalize all input features and pass them to the hidden layer; The hidden layer extracts data features through multi-layer nonlinear mapping: the hidden layer consists of several hidden layers, and the number of nodes in the lth hidden layer is h l , the nodes in the hidden layer process the data in the input layer through a nonlinear activation function to extract the complex patterns and features in the data; the calculation formula of the hidden layer is: in, represents the output of the jth node in the lth layer, represents the output of the i-th node in the previous layer, represents the weight from the i-th node to the j-th node in the l-th layer, is the bias term of the jth node in the lth layer, and f(·) is the activation function; The output layer generates the classification results of the fault mode: it receives the features passed by the hidden layer and converts them into specific fault mode classification results. The number of nodes in the output layer corresponds to the number of fault mode classifications of the system. If the number of classifications is set to C, the calculation formula of the output layer is: Among them, y c is the probability value of the failure mode classification result, indicating the probability that the system belongs to the cth failure mode. is the output of the last hidden layer, is the weight from the last hidden layer to the output layer, is the bias term of the output layer, L is the total number of hidden layers, and c represents the classification category; and the output value is converted into a probability value through the softmax function, the formula is: Among them, z i Represents the input of node i in the output layer.

10. The intelligent DC power supply full-process fault recording and analysis system according to claim 8, It is characterized in that The historical data comparison unit includes: Extract historical fault data features: Extract historical fault data related to the currently identified fault mode from the long-term data accumulation module. The data includes the historical fault mode feature vector H k ={h k1 ,h k2 , ..., h kn }, where k represents the number of the historical fault event, h ki represents the i-th feature associated with the fault event; Calculate the characteristic vector of the current fault mode: Based on the fault mode identified by the current system, calculate its corresponding characteristic vector F = {f 1 , f 2 , ..., f n }, where f i Represents the value of each characteristic parameter in the current fault mode; Similarity calculation: Use the cosine similarity algorithm to calculate the current fault mode feature vector F and each historical fault mode feature vector H k The similarity between k The calculation formula is: Among them, F.H. k represents the vector inner product, ||F|| and ||H k || respectively vectors F and H k The modulus length; similarity S k The value range of is between [-1, 1]. The closer the value is to 1, the more similar the current fault mode is to the historical mode. Risk level assessment: Based on the calculated similarity S k , evaluate the risk level of the current system state, set the similarity threshold θ, if max(S k ] exceeds the threshold θ, the current fault mode is judged to be similar to the high-risk historical mode, and the system is judged to have a high risk; if max(S k ) is close to 1 and exceeds the threshold θ, the risk level is high; if max(S k ) fluctuates around θ, the risk level is medium; if max(S k ) is less than θ, the risk level is low; Output risk assessment results: The assessed risk level results are transmitted to the early warning signal generation unit. If the assessment result is high or medium, a corresponding early warning signal is generated to remind the system operator to take corresponding measures; if the risk level is low, the result is recorded but no early warning signal is triggered.

Images